An optimal moving policy for a mobile agent in distributed networks

An optimal mobile agent moving policy for a mobile agent network where a single mobile agent moves and collects management information, is studied. The network has one management station that maintains many network elements each of which generates the management information (e.g., the updated state of the element). We formulate this network model as a semi-Markov decision process (SMDP) and compute an optimal policy with a policy iteration algorithm. It is shown that an optimal mobile agent moving policy achieved by the policy iteration algorithm has a cost reduction effect for the network in comparison to a heuristic policy.